CN101027694A - Conformal segmentation of organs in medical images - Google Patents
Conformal segmentation of organs in medical images Download PDFInfo
- Publication number
- CN101027694A CN101027694A CNA2005800320441A CN200580032044A CN101027694A CN 101027694 A CN101027694 A CN 101027694A CN A2005800320441 A CNA2005800320441 A CN A2005800320441A CN 200580032044 A CN200580032044 A CN 200580032044A CN 101027694 A CN101027694 A CN 101027694A
- Authority
- CN
- China
- Prior art keywords
- edge
- pixel
- resistivity
- equipment
- image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 230000011218 segmentation Effects 0.000 title claims abstract description 39
- 210000000056 organ Anatomy 0.000 title claims abstract description 8
- 238000000034 method Methods 0.000 claims abstract description 43
- 238000010586 diagram Methods 0.000 claims description 38
- 239000003550 marker Substances 0.000 claims description 10
- 238000001514 detection method Methods 0.000 claims description 5
- 239000012634 fragment Substances 0.000 claims description 5
- 238000004590 computer program Methods 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims 2
- 230000001419 dependent effect Effects 0.000 abstract 1
- 230000006870 function Effects 0.000 description 9
- 230000008859 change Effects 0.000 description 7
- 230000008569 process Effects 0.000 description 6
- 230000009466 transformation Effects 0.000 description 5
- 238000013467 fragmentation Methods 0.000 description 4
- 238000006062 fragmentation reaction Methods 0.000 description 4
- 238000013507 mapping Methods 0.000 description 3
- 230000005540 biological transmission Effects 0.000 description 2
- 238000003709 image segmentation Methods 0.000 description 2
- 230000000877 morphologic effect Effects 0.000 description 2
- 230000008520 organization Effects 0.000 description 2
- 241001270131 Agaricus moelleri Species 0.000 description 1
- UFHFLCQGNIYNRP-UHFFFAOYSA-N Hydrogen Chemical compound [H][H] UFHFLCQGNIYNRP-UHFFFAOYSA-N 0.000 description 1
- 206010028980 Neoplasm Diseases 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 201000011510 cancer Diseases 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000002591 computed tomography Methods 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 210000002216 heart Anatomy 0.000 description 1
- 229910052739 hydrogen Inorganic materials 0.000 description 1
- 239000001257 hydrogen Substances 0.000 description 1
- 210000003734 kidney Anatomy 0.000 description 1
- 210000004185 liver Anatomy 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 210000000496 pancreas Anatomy 0.000 description 1
- 210000002307 prostate Anatomy 0.000 description 1
- 238000001959 radiotherapy Methods 0.000 description 1
- 238000009877 rendering Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000001356 surgical procedure Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/12—Edge-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10072—Tomographic images
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
Abstract
A method and an apparatus for segmenting contours of objects in an image is disclosed. It particularly applies to the segmentation of body organs or parts depicted in medical images. An input image containing at least one object comprises pixel data sets of at least two dimensions. An edge-detected image is obtained from the input image. Markers are selected in the edge-detected image, and assigned respective fixed electrical potential values. An electrical potential map is generated as a solution to an electrostatic problem in a resistive medium having a resistivity dependent on the edge-detected image, with the markers defining electrodes at the respective fixed potential values. Object contours are estimated, for example by thresholding, from the generated potential map.
Description
The present invention relates to the segmentation of image.More specifically, the present invention proposes the process that the robust of segmentation is carried out on an a kind of big class border that is used for different and discrete object that the digital picture of medical image is particularly described.
Digital picture of such segmentation technique processes so as to detect, classification and the discrete object enumerating wherein to be described.It is included in determines contours of objects in the interesting areas, that is, their profile or border, this is useful for the identification and the location of object and shape, form, size, motion of analyzing them or the like.
Segmentation is one of the most challenging task in Flame Image Process, because it need carry out semantic understanding to image to a certain extent.It has also represented a kind of problem of difficulty, because digital picture lacks enough information usually the possible solution of segmentation problem is limited in the little set that has comprised the correct solution of separating.
Image segmentation finds general application in field of medical images, especially computed tomography photography (CT) image, radioscopic image, ultrasonic (US) image, magnetic resonance (MR) image or the like.Special hope positions the various organization objects (such as heart, prostate, kidney, liver, pancreas or the like) that can see in such medical image.Organization object can be used to planning and carry out medical surgery with respect to the position of surrounding, such as operation, be used for radiation therapy of cancer or the like.
Image segmentation is to carry out on the medical image of digital form.Digital picture such as the such target of the part of people's health is a data set that comprises a series of data cells, and each data cell has the digital data value corresponding to target property.This characteristic can be measured on the visual field of imageing sensor in the time period clocklike by imageing sensor.It also can calculate according to data for projection according to pixel grid.The pairing characteristic of data value can be hydrogen richness in the RGB component that separates, X ray attenuation coefficient, MR of light intensity, the coloured image of black-and-white photograph etc.
In known fragmentation technique, Watershed Transformation provides easily and strong instrument mutual and segmentation automatically." the morphologic segmentation " of this classics can think to flood the gradient image that is seen as to topographical surface.At first, make up gradient image.Secondly, for interested object, the mark inside automatically or alternatively stipulating.At last, make up the basin that interrelates with mark, topographical surface is little by little flooded (introducing time parameter).With the shape of mark in the area-of-interest with arrange that irrelevant result can be by obtaining containing the two global minimization of topographical surface and complete mark group.
At " Multiscale Morpho1ogical Segmentation based onWatershed; Flooding and Eikonal PDE (based on the basin, flood the multiple dimensioned morphologic segmentation with mirror image PDE) ", author F.Meyer and P.Maragos, Scale-Space ' 99, LNCS 1682, among the pp.351-362, according to the effective ways that are used for the partial differential equation (PDE) that is called as mirror image equation (Eikonal Equation) is found the solution, one that proposes Watershed Transformation becomes example.It allows to calculate the self-adaptation hypermetric distance that is determined, so that be considered to separate fartherly than the point that is not separated by this edge by an each point that the edge separated.
When such technology is used for the segmentation of medical science organ, when having weak contrast, the object bounds of part can run into restriction.These shortcomings are clearly will be by a plurality of object of segmentation the time when handling.This technology can cause losing the weak edge of the caused leakage in edge, or the segmentation of " leakage ", because some object is not looked at as closed object.Fig. 1 a shows an example with image of two concentric features being represented by they respective edges (through edge-detected image), promptly, star surface 10 and a feature 20 that is comprised in the circle of star surface 10 the insides fully with closed edge.Owing to lack contrast, can not detect the little marginal portion 22 of circular feature 20.The above-mentioned basin technology that is applied to the image of Fig. 1 a causes the result of Fig. 1 b, has only determined surface 10 therein.Really, circular feature 20 is lost, because will be by little marginal portion 22 " leakage " in circular feature 20 for the basin of surface 10 structures.
The fragmentation technique that the purpose of this invention is to provide the above-mentioned shortcoming of avoiding Watershed Transformation of robust.Another object of the present invention provide have that the high precision border determines with the fragmentation technique that can handle a plurality of borders with different contrasts.
Therefore, the invention provides according to the staging device of claim 1 and comprise the medical image system of such equipment.The present invention also provides according to the method for claim 14 with according to the computer program of claim 15.
This segmentation problem is transformed into stable state electrostatic current problem of equal value, and wherein the label fixed current potential is used as boundary condition.The present invention utilizes so-called Conformal Mapping, and promptly a kind of normalized view data self-adaptation Laplce partial differential equation (PDE) separates.It avoids the leakage that is caused by the edge when lose some marginal portions, and causes level and smooth interpolation shape, thereby makes the edge closure when weaken the contrast on border part or do not exist.So, can reach high-precision border and determine, and the fragmentation technique of one or more embodiment of the present invention can be handled a plurality of borders with different contrast.
When the explanation considered in conjunction with the accompanying drawings after this, will understand other characteristic of the present invention and advantage, wherein:
Fig. 1 a is the synoptic diagram of two concentric features in the image,
Fig. 1 b is to use the synoptic diagram of final segmentation of feature of Fig. 1 a of known Watershed Transformation,
Fig. 2 is the process flow diagram according to segmentation method of the present invention,
Fig. 3 is to use according to the synoptic diagram of segmentation method of the present invention according to the current potential mapping of the feature generation of Fig. 1 a,
The synoptic diagram of the final segmentation of the feature of Fig. 1 b when Fig. 4 is to use according to segmentation method of the present invention,
Fig. 5 is the synoptic diagram of the multi-purpose computer of programming in accordance with the teachings of the present invention.
The present invention handles the segmentation of the object on the image.Though the present invention shows that with the software implementation scheme its also available for example hardware component in the graphics card of computer system is realized.
Referring now to accompanying drawing (more particularly Fig. 2), show process flow diagram on the figure according to exemplary segmentation method of the present invention.Total scheme is included in step 100 and initially obtains one comprise will be by the digital medical 2D or the 3D rendering of the object of segmentation.If required, this step can comprise that use file converter components is transformed into another necessary form to image from a file layout.After this input picture that finally obtains is called as M, and p is the pixel label in image.Below, image and its data will represent that therefore, M both had been meant input picture with same title, also be meant the input data of pixel p.After this method also can easily be generalized to the image of dimension greater than 3D.
In second step 110, input picture M is carried out and handles so that provide one through edge-detected image ED (p).Through edge-detected image ED (p) is that method known on by operation technique is created, and it is such as local variance method or gradient intensity method.Original input data M is subjected to rim detection, like this, determines edge strength data ED (p) thereby its edge is carried out to detect, so that distinguish fringe region and other zone of an object.Alternatively, input picture M can at first carry out the edge vividness by suitable technology and feature strengthens, and has the image that strengthens vividness with generation.
At the pixel value ED (p) in edge-detected image feature among the ROI is described.Their representation feature protrusion amounts, the latter can be a pixel intensity value, or any suitable data relevant with characteristic strength in the image M.Partial gradient in pixel intensity is in the most typical amount that will be used in edge-detected image.
The different object that comprises certain number through edge-detected image ED (p).In explanation after this, unless propose in addition, will be as an illustration with two objects.Those skilled in the art can be generalized to big number to it at an easy rate.
Can be modified to an edge strength data that (that is, unlikely are the place of organ contours) in addition at area-of-interest (ROI) through edge-detected image ED (p) and be set to zero.
For by means of finding the solution electrostatic problem, each pixel p through edge-detected image ED (p) or ROI is stipulated a potential function phi object fragments.According to the conversion of segmentation method use of the present invention from the segmentation problem to the electrostatic problem, separating of it is potential function phi (p).
In step 120, going up selected marker through edge-detected image ED (p).Fixing potential value be defined be attached to different marks on.These potential values are used as the fixing boundary condition of electrostatic problem.
For each object that is identified, need inner marker Min and external label Mout through edge-detected image ED (p).These inside and outside marks can be looked at as with to inside and outside border that each object interrelates.For each object, inner boundary can only limit to the point Anywhere in this object the inside, and outer boundary can be the external edge edge of image, or is in the space between this object and the adjacent object.Inside and outside mark can be input to computer system by user (for example, by mouse or similar pointing apparatus), or is provided with automatically.The existing knowledge of relevant extraction shape can be by providing expection the inside and outside mark general shape of object outline be introduced in according to method of the present invention so that algorithm robust more.
For each object, the mark Min of the inside treats as the electrode with current potential Vin, and similarly, the mark Mout of outside treats as the electrode with different current potential Vout.The numerical value of current potential Vin and Vout is arbitrarily, for example can be-100 volts and+100 volts.
In step 130, use ED (p) to generate resistivity map, that is, resistivity distribution (or, equivalently, distribution of conductivity) be mapped on edge-detected image ED (p): each pixel p receives a resistivity value ρ (p) according to it at the numerical value in edge-detected image.Typically, resistivity value can equal the pixel intensity value in edge-detected image, has higher numerical value in the medium that ratio on the target edges of hypothesis is being got involved.Dielectric electricalresistivity (p) is taken as the increasing function of edge intensity value computing ED (p) usually, and it will not get the numerical value less than positive minimum value.
In a preferred embodiment, resistive medium conductivity gamma (p)=1/ ρ (p) (per unit square in 2D, or in 3D per unit cube) is taken as
Wherein A is the parameter that is selected as the magnitude of expection noise level in edge-detected image ED (p), and in other words, the electricalresistivity (p) at the pixel p place is proportional to A
2+ ED (p)
2
In step 140, find the solution the 2D electrostatic problem.Given 2D resistivity distribution and behind potential value Vin, the Vout of marked locations can use known numerical analysis method to find the solution stable state electrostatic current equation, to draw Potential distribution.
In watershed transform technique, " flow problem " solved by flooding the gradient image that is looked at as topographical surface.The partial differential equation of finding the solution (PDE) depends on the time, is progressively to cover full topographical surface because flood.Here, be to find the solution a steady current problems of static state according to the notion of method of the present invention, it is prescribed on the resistive medium that its resistivity and edge-detected image data interrelate.This medium comprises the mark that before had been prescribed as electrode for the boundary condition that is provided with.Step 140 is that each pixel of edge-detected image ED (p) generates potential function phi or the mapping of conformal current potential, with as separating for above-mentioned electrostatic current problem.
Make that J is a current density vector in resistive medium.J (p), Φ (p) and γ (p)=1/ ρ (p) they are the functions of pixel p, as previously mentioned, p refers in initial pictures M or the position in edge-detected image ED (p).Represent the gradient vector operator with , these three amounts interrelate by following relational expression:
J(p)=γ(p).Φ(p) (1)
Because hypothesis does not have source and receiver in area-of-interest (ROI), vectorial J (p) satisfies scattered degree (divergence) condition:
.J(p)=0 (2)
Wherein represents the divergence operator, must find the solution the partial differential equation (PDE) that obtains by the expression formula substitution formula (2) from the J (p) of formula (1).
.(γ(p).Φ(p))=0 (3)
This is a PDE who has comprised a unknown number of the potential function phi in ROI (p), and it is subjected to the restriction of the boundary condition stipulated on inside and outside mark.
In ROI, do not comprise current source or receiver (sink).The Potential distribution of seeking resistive medium can cause the conformal potential diagram wanted.This potential diagram is created in the quick change at object bounds place, and slowly changes between the edge.When part edge is lost, allow the level and smooth current potential that inserts in the neighbours of the marginal portion of losing according to method of the present invention.
Can use the multi-grid algorithm of finding the solution electric problem.For example, can implement method of finite difference.Also can use the method for other prior art.
In some example, can cause separating of potential diagram according to method of the present invention, they are presented at around the inner marker each point place current potential of selecting and can change fast, and this can change seeking of segmentation.A kind of method must be designed such that at the marginal point place of ROI rather than at the inner marker place quick current potential taking place changes.Little inner marker is equivalent to little internal electrode, and may aggravate near undesired quick current potential change such these marks.
According to of the present invention for avoiding near mark undesired quick current potential to change in the preferred embodiment of method of (producing undesired pseudomorphism), implement iterative algorithm, so that obtain more accurate potential diagram.It seems from the viewpoint of reality, after the 5th step 140, that is, after potential diagram generates, use the potential diagram Φ (p) that generates to upgrade conductivity gamma (p) at additional step 145.In a preferred embodiment of the invention, the numerical value γ of modification
m(p) be:
γ
m(p)=B.γ(p)
2.|Φ(p)| (4)
Wherein γ (p) be with generate potential diagram Φ (p) before step in the conductivity used, and B is an arbitrary constant.
This is corresponding to the normalization again of conductivity gamma (p), and carries out on ROI.It allows local relaxation operation, to avoid the potential attraction of some mark.
Identical markers step 12 is held, and is stipulating to have the resistivity γ of renewal
m(p) in the resistive medium behind the electrostatic problem, generated the potential diagram revised with as separating for new electrostatic problem.
In order to converge to stable potential diagram, may need use to carry out iteration several times according to the conductivity of the renewal of (4).
Except in border electrode the inside or the border electrode outside, separating of this potential diagram do not show minimum value or maximal value.Fig. 3 shows the convergent potential diagram that finally obtains, and it can generate according to the feature of Fig. 1 a, and different gray scales is represented each potential value.Shape from two objects of the initial edge-detected image of Fig. 1 a all appears on the potential diagram.The marginal portion 22 of losing is corresponding to the zone of level and smooth potential change.The dark circle 23 at center is corresponding to the zone of wherein having carried out relaxation operation.
In the alternative embodiment by step 146 (selecting non-linear option) expression on Fig. 2, the linear electrostatic PDE of equation (2) can be replaced by NONLINEAR PD E, and wherein conductivity gamma (p) itself is the function of potential diagram Φ (p).This option or be right after in step 130, that is, it is selected that conductivity map generates the back, perhaps after the potential diagram of step 140 generates, that is, is generated during iterative algorithm.
In a preferred embodiment, γ (p) is by nonlinear function γ
N1(p) replace:
When being updated to formula (2), it produces:
This is a NONLINEAR PD E, and it comprises that a conduct is in a unknown number of the potential function phi (p) in the ROI, is subjected to the domination of the boundary condition stipulated on inside and outside mark.
This step back be the 5th step 140 in segmentation method, wherein generate the potential diagram of separating as the non-linear electrostatic problem of (6).
This step is chosen wantonly.It can be right after and generate the back at resistivity map and implement replacing linear PDE (2), and the conductivity of the step 145 that need not describe in the past upgrades, so that the pseudomorphism of also avoiding the quick current potential of electrode to change.
It also can lead to upgrade and be associated with the electricity of step 145, that is, after drawing potential diagram Φ (p), conductivity is updated to γ according to (4) finding the solution the first linear PDE (2)
m(p), the new electrostatic problem that will find the solution then is non-linear PDE (6).Describing identical iterative algorithm with the front is implemented with the use that replaces linear PDE (2) in conjunction with the use of NONLINEAR PD E (6) then.
In step 150, since the conformal potential diagram Φ (p) that calculates in the former step, the profile of the segmentation of generation input picture M.
Conformal potential diagram Φ (p) can obey the dynamic threshold algorithm definition of the threshold value relevant with the position (that is, by).Each the interior pixels p that asks a solution of threshold operation to be included as each object hereto is provided with a neighborhood, and determines the highest and value (this can by suitable histogram technology finish) of which potential value Φ (p) corresponding to the edge strength data according to ED (p) in this neighborhood.When having only a boundary edge to detect, determined that like this this figure of threshold value can provide required object fragments.
When a plurality of organ boundaries need detect, one of these algorithm needs were carried out in the interval that potential phi increases progressively and are dynamically asked threshold step.
Also can use other to ask threshold method, the method that is used for Watershed Transformation that provides in the paper of for example former F.Meyer that mentions and P.Maragos.
Alternatively, different organ boundaries can be stipulated according to priority.For example, first profile that use to allow the regulation the inside that has the segmentation method of an inner marker.Then, this innermost profile can be used as new inner marker and use to be used for second of segmentation method, or the like.
Fig. 4 is given in and asks threshold operation to be applied to the later segmentation profile of potential diagram shown in Figure 3.Two objects from Fig. 1 are that star object 10 and circular object 20 all are identified.At this moment when finished the marginal portion of losing 22 of Fig. 1 a, circular object was exactly the object of a closure now.
About the demonstration of the image of segmentation, the technology that is known in the art can be used to watch the result.For example, when demonstration classifies as all pixels of object the inside of certain gray level, simultaneously be classified as all pixels of described object outside be arranged to before another very different grey level of grey level, thereby profile is become clearly.
Overall computation complexity according to method of the present invention is low.Proposed to use the very effective solution instrument from coarse to meticulous of the electrostatic problem of the approximate hierarchy of discrete circuit.The iteration (being equivalent to about 10 processes) that needs limited number of times for convergence for each internal point of carrying out local relaxation operation.
The present invention also is provided for the equipment of the object fragments in the image, and it comprises deriving means, is used for receiving the input picture that comprises at least one object, and this image comprises the pixel data group of bidimensional at least.Also comprise treating apparatus according to equipment of the present invention, it is used for implementing according to method of the present invention, so that provide edge-detected image according to input picture, and generates the conformal potential diagram of separating as the electrostatic problem that is transformed into from segmentation problem.This equipment also comprises selecting arrangement, is used for being chosen at least one mark on edge-detected image, and mark is associated with the rest potential numerical value of fixing.
The present invention can use the traditional common digital machine or the microprocessor of programming according to the application's instruction to implement easily.This equipment is the part of medical image system advantageously.
Fig. 5 is the block diagram according to computer system 300 of the present invention.Computer system 300 can comprise CPU (CPU (central processing unit)) 310, storer 320, input equipment 330, I/O transmission channel 340 and display device 350.Miscellaneous equipment connects as additional disk drives, annex memory, network ... or the like also includable, but be not expressed.
Claims (15)
1. be used for the equipment of the object fragments of image is comprised:
-input media is used for receiving the input picture that comprises at least one object (10,20), and described image comprises the pixel data group of two dimension at least;
-be used for from the device of described input picture generation through edge-detected image;
-be used for being chosen in described through the mark of edge-detected image and the device of the potential value that is used for fixing accordingly to described mark appointment;
-calculation element is used for generating a potential diagram with as to depending on that mark wherein comes the regulation electrode with corresponding fixing potential value through the separating of the resistive medium electrostatic current problem of the resistivity of edge-detected image having; And
-be used for estimating the device of object outline according to described potential diagram.
2. the equipment that requires as in claim 1, comprise that also the potential diagram that is used to use generation upgrades the device of the resistivity of described resistive medium, wherein calculation element is arranged to generate an other potential diagram with separating as electrostatic current problem new in the resistive medium to the resistivity after having renewal.
4. the equipment that requires as in claim 1, wherein said is to comprise in the area-of-interest of described input picture of described object to be prescribed through edge-detected image.
5. the equipment that requires as in claim 1, the device that wherein is used for selected marker is arranged to be chosen at least one inner marker and at least one external label in described object outside of described object the inside, the potential value that described inner marker designated first is fixing, and the fixing potential value of described external label designated second.
6. the equipment as requiring in claim 1 is wherein stated resistivity in the pixel p place and is proportional to A
2+ ED (p)
2, wherein ED (p) expression is to the value through the data of rim detection of described pixel p, and A is the constant in the magnitude of expection noise level in edge-detected image.
7. as the equipment of requirement in claim 1, wherein said electrostatic current problem has following form:
.(γ(p).Φ(p))=0
Wherein Φ (p) is that γ (p) is a medium conductivity at the described potential diagram of the pixel p place of described initial pictures regulation, i.e. the inverse of resistivity, and represents the gradient vector operator, and . represent the divergence operator.
8. the equipment as requiring in aforementioned claim is wherein stated resistivity in the pixel p place and is proportional to A
2+ ED (p)
2, wherein ED (p) expression is to the value through the data of rim detection of described pixel p, and A is the constant in the magnitude of expection noise level in edge-detected image.
9. as the equipment of requirement in claim 1, wherein said electrostatic current problem has following form:
Wherein Φ (p) is that γ (p) is a medium conductivity at the described potential diagram of the pixel p place of described initial pictures regulation, i.e. the inverse of resistivity, and represents the gradient vector operator, and . represent the divergence operator.
10. the equipment as requiring in aforementioned claim is wherein stated resistivity in the pixel p place and is proportional to A
2+ ED (p)
2, wherein ED (p) expression is to the value through the data of rim detection of described pixel p, and A is the constant in the magnitude of expection noise level in edge-detected image.
11. as the equipment that requires in claim 3, wherein new electrostatic current problem has following form:
Wherein Φ (p) is that γ (p) is a medium conductivity at the described potential diagram of the pixel p place of described initial pictures regulation, i.e. the inverse of resistivity, and represents the gradient vector operator, and . represent the divergence operator.
12., wherein state resistivity and be proportional to A in the pixel p place as the equipment that in aforementioned claim, requires
2+ ED (p)
2, wherein ED (p) expression is for the value through the data of rim detection of described pixel p, and A is the constant in the magnitude of expection noise level in edge-detected image.
13. a medical image system comprises the device that is used to obtain the input picture of describing organ, and as being used for of requiring at each of aforementioned claim as described in the object (10,20) of the input picture equipment that carries out segmentation.
14. one kind is used for the method for the object fragments of described input picture be may further comprise the steps:
-receiving the input picture (100) that comprises at least one object (10,20), described image comprises the pixel data group of two dimension at least;
-generate through edge-detected image (110) according to described input picture;
-be chosen in described mark in edge-detected image and be used for the potential value that appointment is fixed accordingly to described mark (120);
-generating potential diagram with as for having separate (140) of depending in the resistive medium of the resistivity of edge-detected image (130) electrostatic current problem, mark wherein comes the regulation electrode with the potential value of fixing accordingly; And
-estimate object outline (150) according to described potential diagram.
15. a computer program is used for carrying out at the processing unit of image processing equipment, this computer program comprises the instruction that is used to carry out according to the segmentation method of claim 14 when program product moves in processing unit.
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
EP04300617 | 2004-09-22 | ||
EP04300617.0 | 2004-09-22 |
Publications (2)
Publication Number | Publication Date |
---|---|
CN101027694A true CN101027694A (en) | 2007-08-29 |
CN100583149C CN100583149C (en) | 2010-01-20 |
Family
ID=35677489
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN200580032044A Expired - Fee Related CN100583149C (en) | 2004-09-22 | 2005-09-20 | Conformal segmentation of organs in medical images |
Country Status (7)
Country | Link |
---|---|
US (1) | US7873195B2 (en) |
EP (1) | EP1794717B1 (en) |
JP (1) | JP4747175B2 (en) |
CN (1) | CN100583149C (en) |
AT (1) | ATE389218T1 (en) |
DE (1) | DE602005005348T2 (en) |
WO (1) | WO2006033077A2 (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107872963A (en) * | 2014-12-12 | 2018-04-03 | 光学实验室成像公司 | System and method for detecting and showing intravascular feature |
Families Citing this family (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
DE102006059383A1 (en) * | 2006-12-15 | 2008-06-19 | Siemens Ag | Method for producing resulting pictures of investigation object, involves detecting multiple magnetic resonance images of investigation object by magnetic resonance photograph mechanism |
US9558568B2 (en) | 2014-06-27 | 2017-01-31 | Siemens Healthcare Gmbh | Visualization method for a human skeleton from a medical scan |
CN110764087B (en) * | 2019-10-15 | 2021-08-31 | 中国科学院国家空间科学中心 | Sea surface wind direction inverse weighting inversion method based on interference imaging altimeter |
CN112150465B (en) * | 2020-11-24 | 2022-05-13 | 数坤(北京)网络科技股份有限公司 | Blood vessel naming method and device |
Family Cites Families (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5848198A (en) * | 1993-10-08 | 1998-12-08 | Penn; Alan Irvin | Method of and apparatus for analyzing images and deriving binary image representations |
DE69630935T2 (en) * | 1995-09-29 | 2004-11-04 | Koninklijke Philips Electronics N.V. | Image processing method and apparatus for automatically detecting areas of a predetermined type of cancer in an intensity image |
US6047090A (en) * | 1996-07-31 | 2000-04-04 | U.S. Philips Corporation | Method and device for automatic segmentation of a digital image using a plurality of morphological opening operation |
DE69724906T2 (en) * | 1996-10-16 | 2004-07-22 | Koninklijke Philips Electronics N.V. | Numerical image processing method for the automatic extraction of band-shaped objects |
US5859891A (en) * | 1997-03-07 | 1999-01-12 | Hibbard; Lyn | Autosegmentation/autocontouring system and method for use with three-dimensional radiation therapy treatment planning |
JPH11272865A (en) * | 1998-03-23 | 1999-10-08 | Mitsubishi Electric Corp | Method and device for image segmentation |
FR2776798A1 (en) * | 1998-03-24 | 1999-10-01 | Philips Electronics Nv | IMAGE PROCESSING METHOD INCLUDING STEPS OF SEGMENTATION OF A MULTIDIMENSIONAL IMAGE AND MEDICAL IMAGING APPARATUS USING THE SAME |
DE60207569T2 (en) * | 2001-06-26 | 2006-08-17 | Koninklijke Philips Electronics N.V. | ADJUSTMENT OF THE ENERGY FUNCTION IN CALCULATION IN REAL TIME OF CONTOURS FOR PICTORIAL SEGMENTATION |
US6985612B2 (en) * | 2001-10-05 | 2006-01-10 | Mevis - Centrum Fur Medizinische Diagnosesysteme Und Visualisierung Gmbh | Computer system and a method for segmentation of a digital image |
GB0130210D0 (en) | 2001-12-18 | 2002-02-06 | Caladrius Ltd | Segmentation of images using the watershed method |
JP3712234B2 (en) * | 2002-03-19 | 2005-11-02 | 株式会社日立製作所 | Region of interest extraction method and image processing server |
TW554629B (en) | 2002-03-22 | 2003-09-21 | Ind Tech Res Inst | Layered object segmentation method based on motion picture compression standard |
JP3946590B2 (en) * | 2002-07-16 | 2007-07-18 | 富士通株式会社 | Image processing method, image processing program, and image processing apparatus |
US7460709B2 (en) * | 2004-01-23 | 2008-12-02 | Siemens Medical Solutions Usa, Inc. | System and method for multi-label image segmentation |
-
2005
- 2005-09-20 EP EP05784944A patent/EP1794717B1/en not_active Not-in-force
- 2005-09-20 AT AT05784944T patent/ATE389218T1/en not_active IP Right Cessation
- 2005-09-20 JP JP2007531945A patent/JP4747175B2/en not_active Expired - Fee Related
- 2005-09-20 DE DE602005005348T patent/DE602005005348T2/en active Active
- 2005-09-20 US US11/575,701 patent/US7873195B2/en not_active Expired - Fee Related
- 2005-09-20 WO PCT/IB2005/053094 patent/WO2006033077A2/en active IP Right Grant
- 2005-09-20 CN CN200580032044A patent/CN100583149C/en not_active Expired - Fee Related
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107872963A (en) * | 2014-12-12 | 2018-04-03 | 光学实验室成像公司 | System and method for detecting and showing intravascular feature |
CN107872963B (en) * | 2014-12-12 | 2021-01-26 | 光学实验室成像公司 | System and method for detecting and displaying intravascular features |
Also Published As
Publication number | Publication date |
---|---|
US20080137953A1 (en) | 2008-06-12 |
ATE389218T1 (en) | 2008-03-15 |
EP1794717A2 (en) | 2007-06-13 |
WO2006033077A2 (en) | 2006-03-30 |
JP2008513089A (en) | 2008-05-01 |
US7873195B2 (en) | 2011-01-18 |
DE602005005348T2 (en) | 2009-03-26 |
CN100583149C (en) | 2010-01-20 |
DE602005005348D1 (en) | 2008-04-24 |
WO2006033077A3 (en) | 2006-08-24 |
JP4747175B2 (en) | 2011-08-17 |
EP1794717B1 (en) | 2008-03-12 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Qin et al. | Superpixel-based and boundary-sensitive convolutional neural network for automated liver segmentation | |
Suzuki et al. | Computer‐aided measurement of liver volumes in CT by means of geodesic active contour segmentation coupled with level‐set algorithms | |
Panigrahi et al. | Ultrasound image segmentation using a novel multi-scale Gaussian kernel fuzzy clustering and multi-scale vector field convolution | |
Gu et al. | Automated delineation of lung tumors from CT images using a single click ensemble segmentation approach | |
JP2002369079A (en) | Method for improving illness diagnosis by using contrast emphasis presentation | |
Egger | PCG-cut: graph driven segmentation of the prostate central gland | |
Hardisty et al. | Quantitative characterization of metastatic disease in the spine. Part I. Semiautomated segmentation using atlas‐based deformable registration and the level set method | |
Drouyer et al. | Sparse stereo disparity map densification using hierarchical image segmentation | |
Basavaprasad et al. | A comparative study on classification of image segmentation methods with a focus on graph based techniques | |
CN100583149C (en) | Conformal segmentation of organs in medical images | |
Kéchichian et al. | Automatic 3D multiorgan segmentation via clustering and graph cut using spatial relations and hierarchically-registered atlases | |
Shahedi et al. | Deep learning-based three-dimensional segmentation of the prostate on computed tomography images | |
Chen et al. | Incorporating CT prior information in the robust fuzzy C-means algorithm for QSPECT image segmentation | |
Piao et al. | Segmentation of cysts in kidney and 3-D volume calculation from CT images | |
Yao et al. | Mixed spine metastasis detection through positron emission tomography/computed tomography synthesis and multiclassifier | |
Zheng et al. | Automatic liver lesion segmentation in CT combining fully convolutional networks and non-negative matrix factorization | |
Nakhmani et al. | MRI brain tumor segmentation and necrosis detection using adaptive sobolev snakes | |
Suzuki et al. | Interactive segmentation of pancreases from abdominal CT images by use of the graph cut technique with probabilistic atlases | |
Preedanan et al. | Urinary stones segmentation in abdominal X-ray images using cascaded U-net pipeline with stone-embedding augmentation and lesion-size reweighting approach | |
Jiang et al. | A hybrid method for pancreas extraction from CT image based on level set methods | |
Xu et al. | Connecting Markov random fields and active contour models: application to gland segmentation and classification | |
Audette et al. | Towards patient-specific anatomical model generation for finite element-based surgical simulation | |
Baek et al. | Cost aggregation for stereo matching using total generalized variation with fusion tensor | |
Tsou et al. | Anatomy packing with hierarchical segments: an algorithm for segmentation of pulmonary nodules in CT images | |
Tao et al. | A quantitative evaluation of lung nodule spiculation based on image enhancement |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20100120 Termination date: 20180920 |